The OpenClaw ecosystem faces ongoing security vulnerabilities, including a recently patched one-click remote code execution exploit that allowed attackers to hijack AI agents through malicious web pages. The exploit chain leveraged cross-site WebSocket hijacking due to missing origin header validation, enabling attackers to steal authentication tokens and execute arbitrary code in milliseconds. Additionally, Moltbook, an OpenClaw-adjacent social network for AI agents, exposed its database publicly with accessible API keys, potentially allowing attackers to impersonate high-profile AI agents. Both issues have been addressed, but researchers continue discovering security flaws in the rapidly evolving ecosystem.
307 Impressions